Sunday, February 9, 2014

Clicker Expo 2014: Ken Ramirez: Who Nose? (scent detection overview)

I love Ken Ramirez. Everyone loves Ken Ramirez.

in 1997 Ken was brought in to provide an R+ perspective for SAR group, now consults with many SAR and law enforcement agencies
discovered many myths and a focus and reliance on less important aspects of task

types of scent detection: SAR (general and specific), tracking, explosives, drugs, agriculture detection/airports, arson detection, medical detection, wildlife detection (conservation and invasive species, including home pests), sport (hunting, obedience, nose work), and more!

basic training sequence:
train/determine alert or indication behavior
present the scent, cue alert behavior
short search for scent
dilute strength of scent
disguise scent
long search for scent
search for scent with distractions and obstacles

teaching scent detection
we don't teach it - dogs are already expects at this
we teach: what to find, where to find it, and how to react after finding it
easiest part to teach; only 10-20% of actual work

most complex part, 80-90% of work
proofing - distraction training - obstacle work
once alert on scent is well established, majority of time is spent preparing dogs for real world scenarios: adverse conditions/distractions/obstacles, lots of people in panic, other search dogs
train alert with great reliability first
set dog up for success in early stages so he succeeds and task remains enjoyable

alert of indicator behavior:
aggressive: paw at, dig
passive: sit and stare, down
report: bark, rope pull

the "all clear":
a behavior that indicates the absence of target scent in search area
not commonly used, but something Ken strongly advocates
virtually eliminates false alerts
allows animal to get R+ at completion of every requested search
teach indicator for "all clear" first
do short empty search (ie a few boxes), when dog passes them all, cue down (or other "all clear") and R+ (pair completion with new cue)
expand size of search area
alternate real hides/searches with clean area searches
same R+ for all clear as for alert

brownie analogy: if you train a dog to find chocolate using a browie, he may alert on sugar or flour or oil as well
container - inadvertently teaching dog to alert on plastic container, baggie, tape, etc - changing the container regularly
mixed scents - dog learns companion scent
inadvertently touching "clean" object with hands contaminated with scent

pressure to succeed creates anxiety for trainer, increases anxiety in dog
real world scent detection may be serious work, but dog should look at it as fun

false alerts
serious issue in many disciplines - more common with high pressure to succeed or more punishment in training
missed finds even more serious problem - missing the explosive or not finding the victim in time
teach "all clear" signal
can be caused by handler error - accidental cuing

ongoing training is necessary
mimic real world conditions - long drive, middle of the night practice, panic

common errors:
if food is used, has to be faded and actual scent used as fast as possible
use of corrections - almost all scent detection trainers use R+ after correct finds, but many traditional trainers insist on using correction to prevent reactions, punish mistakes, maintain "obedience" - causes shutting down and mistakes
forward chaining - shaping alert last - less reliable
using modeling - shaping alert by pushing butt down - less reliable, slower learning, dog refocuses on you instead of the scent
micromanagement - prompting (because trainer knows location); inadvertently teaching to alert for what we are doing and not what they are smelling; talking to the dog
not trusting the dog - once trained and reliable, never doubt your dog when showing normal alert behavior; causes accurate but unknown finds to not be R+
showing the find before R+ - need immediate R+

(Take that, K9 Nose Work.)


  1. How interesting about the "all clear" signal. I like it a lot! Thanks for sharing. I will take this into consideration when I begin training for NW3, which hopefully isn't that far off (our NW2 trial is one week from today - GAHHH!)

    I completely agree with forward chaining being the least reliable way to train - it's such a default for so many trainers. I'm glad I'm not alone. :)

    1. Glad you found the notes useful!

      Ken's emphasis on backchaining/teaching the alert first is also what I learned from a Steve White lecture some years back, and it's also how it's taught at Denise Fenzi's online academy. It made more sense to me right away. You're definitely not alone! I have K9NW friends who are devoted to the forward-changing method, though.